Thinking:
Compared to the full dataset, my selected tracks (wednesday-w-1 and wednesday-w-2) exhibit above-average danceability, indicating a strong groove and rhythmic presence. However, they are not the most danceable—that distinction belongs to roemer-i-2, which sits at the top of the chart. Interestingly, although my tracks have a higher tempo, they are still outperformed in danceability. This suggests that tempo alone does not determine how danceable a track feels. There might be other influential features at play, such as engagingness, valence, or instrumental texture.
Thinking:
While tempo and approachability showed little or no relationship to danceability, arousal appears to have a weak-to-moderate positive correlation with it. The plot shows a general upward trend, indicating that more energetic or intense tracks tend to be slightly more danceable. However, the correlation is not particularly strong—many low-arousal tracks still have high danceability, and vice versa. This suggests that while energy might help, it is not the sole or strongest factor in determining how danceable a track is.
Thinking:
Although it might seem intuitive that more approachable tracks are also more danceable, this plot reveals no clear linear relationship between the two features. Songs with similar approachability scores exhibit a wide range of danceability values, suggesting that danceability is not primarily driven by affective features like approachability. Taken together with the previous hypotheses involving tempo and arousal, this finding leads to a broader conclusion: Track-level descriptors alone may not sufficiently explain why a piece of music feels danceable. Instead, as discussed in music structure analysis (Chapter 4), danceability may emerge from deeper mid-level features, such as: Chroma-based features, which capture harmonic and melodic repetition, MFCC-based features, which reflect timbral contrast, and tempograms, which encode rhythmic energy over time. In short, danceability is not simply a sum of isolated attributes, but a result of how musical elements interact temporally and structurally — across sections, grooves, and repetitions. Therefore, I decided to shift my focus from the entire corpus to a deeper structural analysis of my own selected track. By visualizing features such as chromagrams and MFCCs over time, I aim to uncover whether my track contains structural signatures that contribute to its danceability.
This track exhibits clear harmonic repetition, with strong vertical striping indicating repeated pitch patterns. This kind of structure is closely related to groove and may contribute to its danceability.
Track 1 shows stable timbral energy concentrated in lower MFCC coefficients, suggesting a consistent and smooth sonic texture that supports groove.
Track 1 reveals a highly regular harmonic structure, as shown by the dense and symmetrical grid-like patterns in its chroma-based self-similarity matrix. This indicates strong harmonic repetition—such as loops or consistent chord progressions—which may support the feeling of groove and enhance danceability.
Track 1 shows clear sectional structure in its timbre-based self-similarity matrix, with large, cohesive blocks indicating repeated segments. These repeating timbral sections likely reflect a structured form (e.g., A–B–A–C), enhancing rhythmic predictability and supporting groove and danceability.
Compared to Track 1, this track shows more harmonic variation but less repetition. It might be more expressive or fluid, but less predictable, which may slightly reduce its danceability.
In contrast, Track 2 exhibits more activity in the mid-level MFCCs (especially coefficients 4–6), indicating greater timbral variation across time. While this variation may add sonic interest and subtle rhythmic motion, it could also disrupt the regularity needed for a strong danceable groove. Overall, the more homogeneous timbre of Track 1 might contribute to stronger perceived danceability, whereas Track 2’s textural complexity could either enhance or interfere with it depending on how listeners interpret the changes.
Track 2, in contrast, shows a more varied and diffuse similarity structure. While there are some repeated harmonic segments, the overall pattern is less consistent. This may reflect expressive variation or structural shifts that reduce rhythmic predictability and, in turn, slightly weaken the groove.
Track 2, on the other hand, presents a more fragmented and intricate similarity pattern. The lack of strongly defined repeating sections suggests continuous timbral variation, which may offer richness and expressiveness but potentially disrupts the consistency needed for strong perceived danceability.